Multimodal Learning for Just-In-Time Software Defect Prediction in Autonomous Driving Systems

Faisal Mohammad; Duksan Ryu

doi:10.1109/BigComp64353.2025.00024

Multimodal Learning for Just-In-Time Software Defect Prediction in Autonomous Driving Systems

Faisal Mohammad^*
, Duksan Ryu

^*Corresponding author for this work

Jeonbuk National University

Research output: Contribution to journal › Conference article › peer-review

1 Scopus citations

Abstract

The growing importance of reliable software in autonomous driving technologies underscores the need for effective defect prediction. This paper introduces a multimodal learning approach for just-in-time software defect prediction (JIT-SDP) using transformers to integrate diverse data modalities such as code features, metrics, and context. By employing attention mechanisms, the model combines outputs from text and tabular data via fully connected layers to predict defects. Experiments on datasets from Apollo, Carla, and DonkeyCar demonstrate that the approach outperforms state-of-the-art models like CodeBERT and GraphCodeBERT, with accuracy improvements of 7.55% to 17.31%, showcasing its potential to enhance software reliability.

Original language	English
Pages (from-to)	88-89
Number of pages	2
Journal	Proceedings of the IEEE International Conference on Big Data and Smart Computing, BIGCOMP
Issue number	2025
DOIs	https://doi.org/10.1109/BigComp64353.2025.00024
State	Published - 2025
Event	2025 IEEE International Conference on Big Data and Smart Computing, BigComp 2025 - Kota Kinabalu, Malaysia Duration: 2025.02.9 → 2025.02.12

Keywords

Autonomous Driving Software Systems
Just-In-Time Software Defect Prediction (JIT-SDP)
Multimodal Learning
Multimodal Transformers

Quacquarelli Symonds(QS) Subject Topics

Computer Science & Information Systems
Data Science

Access to Document

10.1109/BigComp64353.2025.00024

Cite this

@article{d4490535b9d5431594de513473b7a277,

title = "Multimodal Learning for Just-In-Time Software Defect Prediction in Autonomous Driving Systems",

abstract = "The growing importance of reliable software in autonomous driving technologies underscores the need for effective defect prediction. This paper introduces a multimodal learning approach for just-in-time software defect prediction (JIT-SDP) using transformers to integrate diverse data modalities such as code features, metrics, and context. By employing attention mechanisms, the model combines outputs from text and tabular data via fully connected layers to predict defects. Experiments on datasets from Apollo, Carla, and DonkeyCar demonstrate that the approach outperforms state-of-the-art models like CodeBERT and GraphCodeBERT, with accuracy improvements of 7.55\% to 17.31\%, showcasing its potential to enhance software reliability.",

keywords = "Autonomous Driving Software Systems, Just-In-Time Software Defect Prediction (JIT-SDP), Multimodal Learning, Multimodal Transformers",

author = "Faisal Mohammad and Duksan Ryu",

note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; 2025 IEEE International Conference on Big Data and Smart Computing, BigComp 2025 ; Conference date: 09-02-2025 Through 12-02-2025",

year = "2025",

doi = "10.1109/BigComp64353.2025.00024",

language = "English",

pages = "88--89",

journal = "Proceedings of the IEEE International Conference on Big Data and Smart Computing, BIGCOMP",

issn = "2375-933X",

number = "2025",

}

TY - JOUR

T1 - Multimodal Learning for Just-In-Time Software Defect Prediction in Autonomous Driving Systems

AU - Mohammad, Faisal

AU - Ryu, Duksan

PY - 2025

Y1 - 2025

N2 - The growing importance of reliable software in autonomous driving technologies underscores the need for effective defect prediction. This paper introduces a multimodal learning approach for just-in-time software defect prediction (JIT-SDP) using transformers to integrate diverse data modalities such as code features, metrics, and context. By employing attention mechanisms, the model combines outputs from text and tabular data via fully connected layers to predict defects. Experiments on datasets from Apollo, Carla, and DonkeyCar demonstrate that the approach outperforms state-of-the-art models like CodeBERT and GraphCodeBERT, with accuracy improvements of 7.55% to 17.31%, showcasing its potential to enhance software reliability.

AB - The growing importance of reliable software in autonomous driving technologies underscores the need for effective defect prediction. This paper introduces a multimodal learning approach for just-in-time software defect prediction (JIT-SDP) using transformers to integrate diverse data modalities such as code features, metrics, and context. By employing attention mechanisms, the model combines outputs from text and tabular data via fully connected layers to predict defects. Experiments on datasets from Apollo, Carla, and DonkeyCar demonstrate that the approach outperforms state-of-the-art models like CodeBERT and GraphCodeBERT, with accuracy improvements of 7.55% to 17.31%, showcasing its potential to enhance software reliability.

KW - Autonomous Driving Software Systems

KW - Just-In-Time Software Defect Prediction (JIT-SDP)

KW - Multimodal Learning

KW - Multimodal Transformers

UR - https://www.scopus.com/pages/publications/105006625729

U2 - 10.1109/BigComp64353.2025.00024

DO - 10.1109/BigComp64353.2025.00024

M3 - Conference article

AN - SCOPUS:105006625729

SN - 2375-933X

SP - 88

EP - 89

JO - Proceedings of the IEEE International Conference on Big Data and Smart Computing, BIGCOMP

JF - Proceedings of the IEEE International Conference on Big Data and Smart Computing, BIGCOMP

IS - 2025

T2 - 2025 IEEE International Conference on Big Data and Smart Computing, BigComp 2025

Y2 - 9 February 2025 through 12 February 2025

ER -

Multimodal Learning for Just-In-Time Software Defect Prediction in Autonomous Driving Systems

Abstract

Keywords

Quacquarelli Symonds(QS) Subject Topics

Access to Document

Other files and links

Fingerprint

Cite this